David Gassko Chat - Post chat discussion Options

[DSG]

Are you satisfied with how park factors are currently calculated and applied?  Assuming that your answer is probably something between yes and no, can you explain how you think their calculation and application can be improved?

Ooh, this is a tough one. Park factors are done about 90% right even in their simplest form. The issue is that those last ten percent are very difficult to make up. I had a few thousand word discussion with some of the best forecasters in the business about park factors here: http://www.hardballtimes.com/main/article/...ndtable-part-2/

I think Mitchel Lichtman’s comment was the best:

“Park adjustments can be tricky and unreliable for many reasons, but they can be done, if you do them right. Some of the keys for doing good park adjustments are using multiple year data and the right regressions, and not using the “multiplicative method” that you mention. There are other “keys” which I am not at liberty to divulge (until 10 years after my death).”

I’m working on my own set of projections (that’s the first time I’ve announced that, I think…another SoSH exclusive ), and the park factors are a bitch. First of all, you cannot use the multiplicative method (which is what everyone uses, i.e. a player will hit 20% more home runs in Coors). The additive method (a player will hit 4 more home runs a year in Coors) is closer, and closer still is the additive method if you’re using batted ball information (a player will hit .003 more home runs per fly ball in Coors). The totally correct way is using something called the odds-ratio method, but that’s a bitch.

Park factors also have to be properly regressed, a different amount for each component, and its better to regress to an expected park factor based on the park’s dimensions, altitude, location, lighting, etc. than it is to zero. But of course calculating that expected park factor is very difficult as well.

Then you have the issue of left/right park factors. I’ve come to the conclusion that the difference between left/right park factors isn’t terribly significant, and that you might be better off lumping them together to increase your sample size. On the other hand, I also don’t quite have all the data I would like to calculate better left/right park factors: I suspect part of the reason the difference is not particularly significant is that in a park like Fenway, left-handed hitters will actually try to hit more fly balls into left field. So they might end up with about the same home run park factor as right-handed batters, but a lot more fly outs or something like that.

And of course, you have to choose what data to include in your sample and what not to include. For example, the Bill James Handbook discards data from interleague games in its park factors. That makes the sample somewhat smaller, but it might make the park factors better.

There’s just a lot of stuff going on when it comes to calculating a *perfect* park factor, though the thing is that even a very basic one (like what Baseball Reference has) will get you most of the way there.

This post has been edited by DSG: Dec 1 2006, 03:19 PM

[DSG]

Are UZR ratings available yet based on 2006 stats? I'm curious know how many of the Gold Glove winners were actually among the best rated at their position. Do you agree that the best defensive players don't always get the award?

UZR ratings have not been publicly available since 2003, but Mitchel has calculated the 2006 UZR numbers.

The best players don’t get the award probably half the time. My winners probably would have been: I-Rod, Mientkiewicz, Hill, Inge, Uribe, Crawford, Patterson, and Rios in the AL, and Molina, Hatteberg, Valentin, Rolen, Everett, Roberts, Beltran, and Giles in the NL. Two of those guys (Rolen and Beltran) actually won a Gold Glove (to be fair, I also agree with the pitcher selections, though those are kind of difficult to judge).

Adam Everett was especially robbed; he’s the best shortstop since (and maybe including) Ozzie Smith and yet he has no Gold Gloves. That’s complete bull. Omar Vizquel is a very good fielder (especially considering his age), but he’s not Adam Everett.

According to Chris Dial (who doesn’t rate pitchers), this year’s Gold Glove winners were an average of +3 runs above average. My winners averaged +12 runs. Gold Glove winners have access to the same information I do—they just keep making poor decisions. Gold Gloves are pretty meaningless, IMO.

And by the way, it’s a pet peeve of mine, but I think it’s ridiculous that they continue to choose three center fielders instead of a left fielder, center fielder, and right fielder for the Gold Glove. Why can’t they just choose four shortstops, instead of an infielder at each position, in that case?

[DSG]

As an appendix to WTE's question on the various stat categories, I have a related philosophical question:

Do you think that the current generation of new statistics, particularly the BP stats (VORP/WARP), the "plus" stats and Win Shares have become too abstract?  It seems that sometimes these stats deal more with the question of who is better than whom and their relative value, rather than telling us detailed information about the players themselves.

Also, are you bothered that the formula for these and other relativistic stats (such as PECOTA, UZR etc) are based on proprietary formulae that require users to basically "trust" their calculations?

Well isn’t our ultimate interest to know how good a player is?

I think the black box problem can be a serious one, though I wouldn’t say that UZR really is a black box. MGL was pretty clear on what he does with UZR in his original two articles, and you could pretty much re-create it if you wanted to. The issue is that the data to calculate UZR costs five figures, and it would take dozens of hours to program the whole thing. If you came into full Stats or BIS play-by-play data, do you really think you’d do anything with it?

As for PECOTA, it sucks that we don’t really know what Nate does with it, but more because we could probably pitch a lot of ideas for improving PECOTA than anything else. Look, BP is running a business, and I understand that they can’t just give away all the details. You can really only judge PECOTA based on its final results, because it is what it is. And the final results are pretty good.

You know, when Bill James was writing in the 1980s, there was a lot of stuff that could just be expressed in a formula or a simple study. Baseball research has long passed that point, and most things just aren’t that simple anymore. That’s just the reality of it.

[DSG]

The  contents of the annual can be found here:

2007 THT Annual

Here are few questions based on the articles that David wrote.

1.  You mentioned over at BTF that pitchers have a very high success rate after TJ surgery.  Does your study include minor league pitchers?  If so, I'm curious how you judged a successful return.  For a Sox specific example, Delcarmen went from a pretty good A ball starting prospect to potentially a solid late inning releiver.  Is that a success despite his move from a high value role to a lower value role?

If minor leaguers weren't included, would you expect a lost year of playing time during what should be a growth phase to be a bigger impediment to future success than the surgery itself?

2.  Is your scheme for predicting breakout players intended to identify young brreakout major league players or does it include minor leaguers as well?  Would you expect there to be different indicators for minor leaguers vs major leaguers?

And what the hell, toss out a couple of names as teasers.

3.  Are you aware of any research on using medical data to help generate more accurate medium and long term playing time projections or at this point is simple Marcel-like projections with an aging correction good enough?

More specifically to the Sox, JD Drew played ~600 games from ages 26-30 how does that impact the number of games he should be expected to play in his age 31-35 seasons?

I could only look at pitchers with major league experience both before and after the surgery in my study. It’s possible that the results don’t apply to all pitchers because of that issue, though I am not sure that this is the case, either. If Delcarmen throws a hundred innings in the major leagues, that’s a success because for most good A-ball prospects, that just doesn’t happen. Look at it this way: Edwin Jackson out-dueled Randy Johnson in a major league baseball game…and now he’s pretty much done. Just making to the majors for any prolonged period of time is a great accomplishment and a success. Just ask Steve Lomasney.

For the breakout article, I too looked only at major league players. I’d guess the indicators should be pretty similar for minor leaguers, but who knows? Here are the odds of some Sox hitters having a breakout year next season (average for batters is around 13%):

Wily Mo Pena – 37%
Coco Crisp – 9%
David Ortiz – 0%

Only Pena has above average odds among Sox hitters (he’s also fourth among all batters).

(I guess that's not such an exclusive since Gordon Edes somehow got his hands on the numbers in the Globe today:

http://www.boston.com/sports/baseball/reds...ox_nixon_in_07/

Sig Mejdal did some really good research in predicting injuries for the 2005 Bill James Handbook, I believe. Unfortunately, he now works for the St. Louis Cardinals, so anything he does with that is owned by the Cards. I guess Will Carroll might have his own system, though I’m unsure. Predicting injuries is one of the great next frontiers. Actually I had an article up at The Hardball Times yesterday related to that question:

http://www.hardballtimes.com/main/article/...r-after-effect/

Without really having any data to back it up, my guess would be that Drew will average 130 games a year over the next five seasons.

[DSG]

Has the efficiency of positioning increased with all this study of zones (by others)?  And do you have an opinion in regard to the frequency of using shifts?  They seem to be almost solely employed against the greats of the game like Big Papi.  Yet guys like Trot Nixon seem to never hit balls to the left of second base and yet never faced shifts.

I agree that teams probably don’t shift enough. If it works against Papi, it should work against Nixon. Though the one issue there is that opportunity cost for Papi to try to hit it down the third base line is greater than it is for Nixon. If Papi alters his swing, he’s taking away the biggest threat he has—the home run. Nixon is just going for a single to a different field than usual. It’s kind of like the intentional walk, which might be a good play against Ortiz but not Nixon.

I doubt positioning has improved much with the study of zones, mainly because there’s one manager in all of baseball who to my knowledge cares at all about all the statistical research out there. Most managers just don’t care, so no matter how much research we do, they’re just going to go by “The Book” (and not “The Book” written by Tom Tango, Mitchel Lichtman, and Andy Dolphin, unfortunately).

[DSG]

1. In your research for the 100 Most Valuable Pitchers article in the forthcoming Annual, did you reach any conclusions about pitcher usage?  Specifically, 5-man rotations vs 4-man rotations, pitch counts, optimum days rest, the use of "Sunday Pitchers, bullpen usage, etc" 

Do you think the current philosophy of pitcher usage is close to optimum, or will it evolve further, and if so, into what?

This is a difficult question to answer. What I found was that pitchers’ careers seem to be extended quite a bit by the conservative pitch counts and five-man rotations we see today. The authors of “The Book” looked into this with much more detail and found that pitchers performed better on four days rest than three, so much so that a four-man rotation would not be better than a five-man rotation, even though better pitchers would get more innings that way. I tend to agree with that for non-statistical reasons.

But baseball is an always-evolving game. It’s very possible that someone will try something different next year, it will work, and a bunch of copycats will spring up. I have no idea what that idea will be...bullpen-by-committee?

[DSG]

Care to comment on any of the recent specific defensive value debates on SoSH?

For SS, debate rages as to whether Alex Gonz is great and Lugo mediocre, or the two are almost even. Somehow, THE question of how awful Micheal Young is keeps edging into this one.

For 3B, the question is if Lowell is much better than Youk. Some have claimed that Lowell was 2 wins better than Youk over all in 06 due to his defensive excellence (and he starts in a hole due to being worse on O).

For OF, many questions swirl. If Manny as awful as many/most systems suggest? Can drew make up the difference between their bats with defense?

What do you make of guys like Crisp and Wily Mo who seem to be much better in one slot than another (ie, Crisp is great in LF, lousy in CF; Wily Mo is not great in Cf but excreable anywhere else).

EDIT: one letter wrong, and I said the opposite of what I meant...

Gonzalez and Lugo are about even; both are above average fielders. Gonzalez might be a little better. Offensively, Lugo is obviously far superior. Young sucks.

Lowell is a better fielder than Youkilis, but not by too much. I like Youks, personally—I think he’s about an average fielder and Lowell is clearly above average but at an age where he should be declining. The difference between the two is probably half-a-win, maybe a win, certainly not two.

Manny is pretty much terrible. A -15 to -20 fielder, IMO. “Chone” Smith’s projection has Manny and Drew exactly equal in total runs above replacement (both are +33), so yes, I would say that Drew’s defense could make up the difference between their offensive abilities. Of course, people will hate it even if he does because no one will realize that our improved pitching will be due partly to trading Ramirez for Drew (essentially).

Center field is a more difficult position than right (actually, it isn’t. There are just more balls hit into center, so teams put better fielders there, which in turn makes the average center fielder a much better fielder than the average corner outfielder), and a few hundred innings from Wily Mo Pena or Coco Crisp is not going to make me change my position (especially when Crisp was injured last season). I can’t imagine how one outfield position would be much easier than another.

[DSG]

In the absence of a Henderson or a Raines, there is a still a tendency of managers to prefer speedy leadoff hitters with relatively low OBPs (the Crisp, Soriano model) over slower but high OBP hitters (Youkilis Boggs).  Two questions regarding this:

1.  Statistically, how many runs would you estimate that such a decision (say, Coco over Youkilis) costs over the course of a season?

2.  Specifically, how many runs would you estimate it cost the Nationals last year to have Soriano hitting leadoff instead of in a power slot, provided his performance were to be unaffected by such a change?

No more than a few runs (for both your questions). Even if you put together the worst lineup order possible (worst hitters at the top, best at the bottom), it wouldn’t cost you more than one or two wins. Lineup construction is overrated, it’s just annoying when a manager costs his team runs for no reason whatsoever.

[DSG]

There is much discussion lately about the consequences for David Ortiz if Manny Ramirez is traded; it sometimes is said that Ortiz will suffer because Ramirez won't be there to protect him in the lineup.  I invite your comment on that particular question, or on the general issue of whether a player's hitting is affected by the strength of the player hitting behind him.

I suggest you order “The Book” (http://www.insidethebook.com/).

The authors show that protection is largely a myth: In fact, hitters do a little better (though the result is not statistically significant) without protection than they do with protection. (That sentence sounds strangely dirty, doesn’t it?) I wouldn’t worry about Ortiz’s performance if we don’t keep Manny.

[DSG]

Regarding the defensive systems you mention in your Feb 3, 2006 article on THT (and any defensive metric, for that matter):

In your discussion of PMR, you mention that it takes into effect how "hard" a ball is hit. I'm sure that some of the other systems attempt to do something similar... but how?

It's my understanding that, for the most part, the descriptive play-by-play elements of defensive metrics are subjective; i.e., an observer notes whether or not a ball in play is a fly ball, line drive, bloop, etc.

Are there any attempts to quantify how hard a ball is hit? Could a ball in play be timed from the moment it is struck until the moment it hits the ground, to determine how "hard" it was hit? Presumably, two balls that reach the same spot in the OF, in the same amount of time, would share a similar arc, right?

A tangible value that combines location measurement with time measurement would provide fairly reliable data on the difficulty of defensive plays in relation to others, and could be invaluable in the sort of probability studies that are already out there. Yes/no?

Both Baseball Info Solutions and Stats Inc. label every ball in play as “hard,” “medium,” or “soft.” Mitchel Lichtman has always used those definitions in UZR. As I said earlier, ideally, instead of all these somewhat subjective designations, we would simply time how long each ball was in the air and its trajectory. Actually, Greg Rybarczyk does that already for home runs at Hit Tracker Online: http://hittrackeronline.com/.

So…yes.

[DSG]

Based on your reasearch for your series 'Does Size Matter', what do you think we can expect from our little man, Dustin Pedroia?

I think Pedroia will be a Bill Mueller type. Pretty good defensively, doubles power, pretty good batting average, very good walk rate. He’s certainly not the top prospect some thought he was a couple years ago, but I still think he’ll be pretty good. I think .290/.360/.420 next year and .300/.390/.440 at his speak sounds pretty realistic, which is very good for a second baseman.

[DSG]

Interpreting your bump as a call for further questions ...

Another BABIP question, related to my last.  No matter how tight BABIP numbers may seem, the "BA" part of the acronym implies one dimension - hit or
out, with no real value to the actual quality of the contact (an infield single being the same as a 500-foot HR.  So BABIP doesn't really have a direct relationship to run production (or runs allowed for pitchers).  Can you comment on this, and if you haven't already done so in responses to other questions, elaborate on what exactly you feel BABIP can tell us.

BABIP is a very complicated subject. I have an article on the relationship between defensive independent numbers and BABIP that I have to write up for The Hardball Times which comes to some interesting conclusions, and it’s really a subject that can be endlessly researched. What you seem to be suggested has often been discussed, which is to look at slugging average on balls in play instead of batting average.

But here’s the thing: Fly ball pitchers allow less hits on balls in play but more extra base hits on balls in play, so what happens is that they end up with just about the same slugging average on balls in play as groundball pitchers.

BABIP is a skill but there’s so much noise there, and so much evens out, that it’s not a terrible mistake to assume that no pitcher has any skill balls in play whatsoever. It’s not optimal, but it won’t make a huge difference either.

[DSG]

Theo Epstein receives a lot of praise from mainstream and online obserevers alike.  What are the main weaknesses of the Epstein front office?

There aren’t many. The bullpen obviously hasn’t performed too well, but I’ve liked a lot of the bullpen moves the Sox have made. Chad Fox gave us a 4.50 ERA in 2003, then moved on to Florida that same year and gave them a 2.13 ERA. Rudy Seanez had an awful year for us after two very good seasons in a row. Byung-Hyun Kim fell apart for no apparent reason. Foulke has battled injuries the past couple of years. None of these were very predictable events. The Red Sox just haven’t had the upper minor league talent needed to fill in holes in the bullpen that some teams have. I think that’s changing.

The Red Sox have been a little too quick to give up on some players, but who knows how much of that is the fault of the media and talk radio especially. Manny should have been gone years ago (he’s my favorite current Red Sox, BTW, but he really is just terrible in the field), but I’m not in the front office—it’s possible they couldn’t do that without taking on a substantial part of his contract, which would negate the whole reason for trading him in the first place.

Overall, I think Theo has done a tremendous job and if the prospects he’s drafted work out for us (and that’s one thing you just never know), we’ll really start to see the fruits of his labor in the next five years. A scary thought when you remember that Epstein already brought us one World Series victory.

[DSG]

So much of the analysis of free agent signings is based on a presumed average value of an expected marginal win of production and yet everybody realizes that each organization will generate an organization specific amount of revenue for each marginal win.  Given the wide range of revenues that he will probably never truly understand the finances of these private companies, do you think that analysists have gone too far grading the financial impact of player moves with such a blunt instrement as the average value of a marginal win?

Somewhat, yes. But only somewhat. I’m currently working on a big project trying to determine the value of a marginal win, taking into account everything I possibly can. It’s a difficult process, and you’re right, we can never really know for sure, but we sure can get a pretty good idea.

I can’t possibly see how you would defend Carlos Lee’s contract…a guy like Gary Matthews Jr. would be more difficult to evaluate given our various uncertainties.

Nate Silver did some good stuff in “Baseball Between the Numbers” that addresses many of your concerns, and Vince Genarro has too. Vince is writing a book, but he published some of his research on the Hardball Times:

http://www.hardballtimes.com/main/authors/vgennaro/

One warning: Their numbers are going to be off because they both use BP’s WARP. What’s important about their research is not the numbers, but the ideas.

[DSG]

What metrics do you feel are most effective for assessing performance of starting pitching? In conjunction with this, what metrics are best for predicting future pitching performance?

Also, I have never seen linear weights used in conjunction with starting pitchers. Why is this?

Thanks

As I told you at MIT, the reason people don’t use linear weights for pitchers is because (a) Until recently, we haven’t had 1B/2B/3B information for pitchers, and ( b ) While a batter has little impact on his run environment, a pitcher has 100% control over it, so Johan Santana should have a completely different set of linear weights than Jaret Wright. A walk against Santana might be worth .20 runs, against Wright, it might be .40 runs. You could use custom linear weights, but then you need to use BaseRuns to generate that, so why not just calculate component ERA (which is like using linear weights, actually) with BaseRuns?

What metrics are best for assessing starting pitching? I like to use a few years of data, regress each component individually, weight each year appropriately, and well then you have a projection! Or do you mean for assessing value? In that case, I’d obviously use my Pitching Wins system.

This post has been edited by DSG: Dec 1 2006, 03:29 PM

[DSG]

The Red Sox for the last few years have pretty consistency underperformed their runs allowed expectations by just about every projection system I can think of, be it ZiPS, PECOTA, or Marcel.

Why do you think this is?

One theory off the top of my head would include bad luck in the form of injuries, or unforeseeable performance regression, both of which can hit for several years in a row just on chance.

Another theory, and probably the one favored by MGL, would be that the Red Sox just have an unbelievably awful defense(which he said cost the Sox about half a run per game in 2006 relative to average - which is of course, a world of hurt in the long run).

The most damning theory would be that the Red Sox, being a sabermetrically friendly organization, are going to tend to select for players who do well by computerized projection systems, but if the projection systems are systemically missing something, then the Red Sox pitching staff will do the same. Because of this, computer projection systems will always overrate organizations like the Red Sox.

These are just three theories, and of course there are others. Obviously it's likely that there's a multitude of different factors at play, so what I guess I'm looking for is your best guess of the ranking of factors.

This is something that deserves more research than I’m going to give it. I think your third theory is an intriguing one, though obviously I’m a pretty big believer in sabermetric tenets, so I’ll put it last.

I’m not sure you can blame the Red Sox defense. Based on The Hardball Times fielding stats, they’ve been average or better (once you adjust for the left field park factor, which we don’t at THT) each of the past three years. Looking at my UZR files, the Red Sox were -58 runs in 2005 and -20 in 2004. I don’t have 2006 data yet, but obviously that doesn’t jive with our numbers. I trust UZR more, but the THT team-level data is pretty damn good. I don’t really know what to say…the only possibility I can see is that our pitchers have been giving up really easy-to-field balls in play (soft line drives, lazy fly balls).

Bad luck definitely sounds about right. As I said earlier, the Red Sox have had a lot of bad luck with the bullpen, and it hasn’t been better with the starters (Schilling in 2005, Wells, Wakefield, Beckett, and Clement last year). Maybe that’s the issue, but I don’t really know. Maybe we’re just cursed.

[DSG]

How well do you think baseball FOs in general do, at the moment, in calculating the value of defense in the contracts they sign with players? How much range is there between the FOs that do this well and those who do it less well? Are teams better at measuring the D of position players, or the dependence on team D of pitchers for these purposes?

IOW, Is defense still a market ineffiecency to be exploited by the "smarter" teams?

Yes, yes, and yes. Did you see the Carlos Lee contract? Do you think the Astros have any idea that his defense negates his offensive value (essentially)? A lot of teams just don’t buy the defensive metrics, and of those that do, many don’t really pay any attention to them. You might see a lot more teams hiring stat guys, but many just do it because you’re supposed to have one, stick him in a dark office somewhere, and don’t listen to anything he has to say.

Defense is still a huge market inefficiency and a big part of the reason that the A’s can continue to win.

[DSG]

What is a baseball contract option worth? How well could current option pricing models be applied?

So, a team offers a player 4 Years/$40M+Team Option @ $10M (or whatever).

How much is that option worth, from the team's perspective?

How much less valuable is the contract from the player's perspective, as opposed to the inclusion of no option, a mutual option, vesting option, or player option?

Of course, I'm don't mean to ask for an exact $value. I'm more interested in the approach you would take to the issue.

I think options are ridiculous. A player or team will rarely exercise an option unless the player’s market value is much higher (if it’s a team option) than the option or much lower (if it’s a player option). So really, when either side agrees to an option that the other side can exercise, they’re saying, “yeah, you can screw me, it’s alright.”

So I think the option is always a bad thing for the team if it’s a player option and a good thing if it’s a team option. By how much? Depends how much the player over- or under-performs.

[DSG]

Alright, thanks for all the questions guys. If you want to read my articles, they're all located here:

http://www.hardballtimes.com/main/authors/dgassko/

Again, if you’re interested in the Annual, an article outlining the contents is available here:

http://www.hardballtimes.com/main/article/...es-annual-2007/

They just started shipping it out yesterday, so if you order now, it’ll get there soon. If you like the work we do at The Hardball Times, this is best way to support us (and if you can afford to, please buy it through our link and not through Amazon. We make much more money that way).

Gordon Edes must have gotten his hands on an advanced copy, and he mentioned on of my articles in the Annual in the Globe today:

http://www.boston.com/sports/baseball/reds...ox_nixon_in_07/

(It’s about three-quarters of the way down. I know I already mentioned this in the chat, but c'mon, it's freaking cool for me, so give me a break.)

I’ve read the whole thing, and in my obviously biased opinion, it’s awesome. If you got last year’s Annual, this one is a step above (and I loved last year’s), if you didn’t, just imagine the greatest thing that has ever happened to you, and multiply that feeling by ten.

I’ll stick around for a couple hours to answer follow-up questions.

Thanks guys, it’s been fun.

[Frisbetarian]

I would like to once again thank David for the informative responses, and would also like to put in my plug for the 2007 Hardball Times Annual. Last year's annual was outstanding and I'm expecting more of the same. If you are interested in current baseball research and metrics, you will not find a better source.

Now, lets get some follow-up questions.